Culling choices

I tend to only remove the very obviously bad images, when I catch them. Some examples would be an Airplane, or dawn frames. Not much as culling goes, really.

I more depend on my ASI Studio and its rejection settings (Is that AI stuff?) which is usually very mild at the stock settings. Sometimes nothing is rejected after my lightly culling the obvious. I might take 200 to 300 images at a certain exposure length, but throw away the last 40 to 60ish due to “dawn burn”, or maybe neighboring trees interfering. (I don’t cull the overhead wires generally, as another example.)

I like to often just run the bulk of my images after my casual culling and save as a jpg for the web. I usually start at a minimum of 30° and run to destruction at astronomical dawn.

My delete key forgives all sins. No picture = didn’t happen. 😄 My SSD I save my images on was getting critically full, so I deleted 2025’s image files to make more room for this years. The old stuff is just old stuff. Onward through the fog! 😜

I think the effectiveness of useful little refinements like this depend a lot on the system in question and conditions. A small refractor that’s under-sampling probably won’t benefit much. With my 150mm aperture, I’m usually sampling at 0.66” or 0.38” per pixel with a tiny 585 sensor. It takes a lot to make things look good with a sensor that small, it pushes me to the wall every time. I’ve found little improvements like this have a large impact. Bottom line is everyone really should just do the testing and see what works best for them.

One last point and this is something I’m still refining. Combining the low frequency data from the full stack and the high frequency information from the deeply culled stack can (I think) give you the best of both worlds. I’m not sure this would be effective with all classes of objects but what’s nice about it is that no data is wasted…work in progress.

I think people reject way more data than they should. You can take a set of data and test this yourself. Throw it all into the blender using PSF Signal weight, and then do the same thing with incrementally more culled sets. You will likely find that WBPP does a great job weighting even your bad subs low enough that you can still keep a lot of the signal you would have tossed out otherwise, and not really harm the end image.

The things I think people should care more about are leftover motes from bad calibration, legitimate clouds, streaks from objects that are too bright to keep, then maybe a little bit of really bad tracking error and egregious seeing.

This is coming from someone that has to process a ton of data from a ton of telescopes, and do it all in a manner that delivers quality, but also doesn’t make me want to rip the remaining hair I have out of my head. :)

I chopped out an airplane last night. I happened to be watching when Frame 47 came in and wouldn’t you know it… an airplane flew directly through the center of the image. A real Bullseye.

I managed to delete it right away. Probably should have saved it for show-and-tell.

Our Airplanes will get removed in stacking, and I don’t inspect all my frames. But since I caught it ….